“Decision Tree” as a Tool for Analyzing the Psychodiagnostics Data Latent Connections

Background. Machine learning methods turn to be an effective tool for data processing and provide ample opportunities for analyzing the results of psychodiagnostics. At the same time, they have a number of advantages over traditional correlation, factor, regression analysis and allow for a systemic analysis of latent relationships between psychological indicators. Search, analysis, and interpretation of latent indirect relationships between features is one of the key problems in modern psychology. In this regard, the “decision tree” method provides wide opportunities for researchers. Objective. To demonstrate the capabilities of the “decision tree” method in the processing of multidimensional psychodiagnostics data using specific examples. Study Participants. The study involved different groups of respondents for processing of heterogeneous original psychodiagnostics data: parents of preschoolers (496 mothers), 83 pre-adolescent schoolchildren (aged 11–12), 79 adolescent schoolchildren (aged 14–16). Methods. The results of two or more tests were analyzed simultaneously. The study implemented “Parental Attitude Research Instrument” (PARI) by Schafer & Bell, “Analysis of family relationships” by Eidemiller, the Tapping test by Ilyin, the Cattell Culture Fair Intelligence Test, the 12-factor personality questionnaire by Cattell (CPQ), the “Unfinished sentences” test by Michal, the personality test by Rogers & Diamond, and the 14-factor personality questionnaire by Cattell (14 PF). Data processing algorithms were implemented on the DEDUCTOR analytical platform. Results. Using the «decision tree» method, a large number of psychological traits were classified according to their relationship to different target attributes (psychological characteristics). The study demonstrated the possibilities of the proposed data processing algorithm for evaluating the latent connections between psychological characteristics and for building the hierarchical structures corresponding to vertical system data analysis. In particular, the proposed tool allowed for identifying the relationships between individual psychological and psychosocial characteristics of pre-adolescent schoolchildren (aged 11–12); between personality traits and adaptation characteristics of 14–15-year-old teenagers; between parental attitudes and family relationships among preschoolers’ mothers. Conclusions. The “decision tree” method allows for the analysis of latent relationships between heterogeneous psychodiagnostics data by the level of significance of input features, even without analyzing the structure of the “tree” itself. The method makes it possible to identify the range of values of psychological features in which the intrasystemic relationships are more stable. The data analysis demonstrates the influence of IQ indicators as individual psychological characteristics on the results of schoolchildren testing. Verbal and non-verbal intelligence (14-PF questionnaire and Cattell Culture Fair Intelligence Test) appear to be significant indicators in the interrelations of psychological features.